THE IMPACT OF RENEWABLES ON THE GERMAN ELECTRICITY MARKET:

THE IMPACT OF RENEWABLES ON THE GERMAN

ELECTRICITY MARKET:

A time-series analysis based on German day-ahead electricity prices.

Master’s Thesis.

Gert-Jan de Jong

Msc. Finance, Energy Focus Area S2785889

Supervisor: Dr. Raymond Zaal MBA

University of Groningen, The Netherlands

July 2, 2018

Abstract

This paper analyses the impact of an increasing capacity of renewable energy on the German day-ahead electricity prices in the period 2004-2017. As both wind and solar energy are closely related with climate conditions, I estimate the impact of wind velocity and sunshine duration on electricity prices as proposed by Mulder and Scholtens (2013). They focus their study on the impact of both economic and climate conditions on Dutch electricity prices. I find that the daily average wind velocity has an economically significant negative impact on German day-ahead electricity prices. Moreover, I find that this negative impact is increasing over time when modelling actual generation output of both wind and solar electricity.

Key words: electricity, energy capacity, climate, Germany, wind, solar, power, renewable energy

1. INTRODUCTION

manage this, investments in flexible power plants have to be evaluated. A flexible power plant would have to be able to be shut off when demand can be met with renewable energy sources and be turned on when additional supply by conventional sources is needed. Both the puzzle of how to subsidize investments in such flexible power plants and how to financially structure them will need a further assessment. Also, from an investor-perspective, using financial instruments like weather derivatives will enable investors to hedge against price movements that are weather dependent. In 2000, the Renewable Energy Sources Act (Erneuerbare-Energien-Gesetz EEG) was formed to enhance the deployment of renewable energy sources (www.iea.org, 2016). In 2004, a new act replaced the Renewable Energy Sources Act of 2000 to further increase the share of renewable energies. In the period after 2004, this act is replaced several times to further enhance both deployment of renewable energy sources and innovation in renewable energy technologies. Due to limited data availability, I am able to cover the period 2004-2017 in this paper. This is an interesting period of research as the electricity market has changed a lot in this time period due to the growth of renewable energy capacity. Especially wind energy capacity has seen rapid growth in this period. Since wind energy in Germany is already quite mature, I expect to find a great negative impact of wind velocity on the electricity prices. For sunshine duration, I also expect to find a negative impact. This relationship is likely to be less powerful as solar energy capacity is relatively low compared to wind energy capacity. I expect to find a positive relationship between the temperature of river water and electricity prices. Power plants use the river water for cooling purposes. When the water exceeds a certain threshold, power producers are obliged to shut down their plants, resulting in a leftward shift of the supply curve. Assuming that demand is constant, this results in higher prices. By analysing the impact of both weather and economic conditions on daily German day-ahead electricity prices, this paper provides answers to the following research questions:

- What is the impact of an increasing capacity of renewable energy sources on German electricity prices?

- Does the impact of an increasing renewable energy capacity increase over time?

have a negative impact on the total marginal costs of production, and thus, on the electricity price, provided that the merit-order curve will shift to the right. This is equal to a rightward shift in the supply curve while demand remains constant, as is displayed in figure 2. In the long run, the increasing share of renewables may have an ambiguous effect on the electricity price. Lower revenues of conventional power plants are a result of the increase in the deployment of renewables. Klessmann et al. (2008) suggest that in the long run, this may even lead to de-investments in conventional plants, which lowers total capacity in times of low wind or solar power availability. Spot prices may increase due to this lower capacity. Ozdemir, Koutstaal, and Hout (2017) find that the increasing deployment of renewables has a noticeable impact in electricity markets. They state that balancing costs increase to adjust for errors in forecasting the supply of renewable energy. Renewable energy comes with additional transmission costs to reach demand centres, and back-up is needed in periods where the production of renewable energy is low. In this paper, I investigate whether the merit-order effect is stronger than the positive impact an increasing share of renewables may have on electricity prices due to de-investments, balancing costs and additional transmission costs. I find a significant negative relationship between the wind velocity and German day-ahead electricity. Moreover, I find that the coefficient for wind velocity is increasing gradually, suggesting that the German day-ahead electricity price is becoming more dependent on ecological factors like the wind over time. When I regress actual wind energy generation on electricity prices, I also find a negative impact, which is growing over time.

This paper proceeds as follows. In section 2, I provide a theoretical background on the merit-order effect and I provide a literature review on previous research in this merit-order effect. In the third section, I give a description of the data used, including sources and some descriptive statistics. In section 4 and 5, the methodology and the results are provided respectively. The conclusion will be presented in section 6.

2. THEORETICAL FRAMEWORK & LITERATURE REVIEW

Graichen, 2014). In 2017, wind energy capacity has seen its largest growth ever in Germany (www.wind-energie.de, 2018). This is reflected in a rightward shift in the merit-order curve, as displayed in figure 2.

Figure 1. Installed capacity per production type for the period 2015-2018. Source: Entsoe

Figure 1 displays the yearly installed capacity of most of the deployed energy production types in Germany. The figure shows a substantial growth in onshore wind capacity, and a moderate growth in solar energy capacity. Also, the increase in the capacity of gas-powered plants is very clear in the period 2016-2018.

Following German policy, renewables have priority supply to the network where electricity is transported from power plants, wind turbines and solar panels to consumers, the so called “grid”. This implies that all low-cost electricity from renewable energy sources are exploited first. Hereafter, the non-renewable energy sources are used, up to the point where electricity supply equals demand. At this point of the merit-order curve, the electricity price is determined as displayed in figure 2.

Figure 2. The merit-order effect. An increase in renewable capacity lowers electricity prices. The y- and x-axes represent the marginal costs in EUR/MWh and the capacity in MW, respectively.

When renewable energy capacity increases from point D to point E, the merit-order curve shifts to the right. With constant demand, this results in an electricity price decrease from Pold to Pnew,1 or to Pnew,2, depending on the elasticity of demand.

When demand is inelastic (D2), the demand curve is relatively steep. An increase in renewable capacity then results in a move from point A to point C in figure 2. Relative to a more elastic demand (D1), this leads to a higher drop in the electricity price. The merit-order effect gives rise to the first hypothesis being tested in this paper:

1. H0: An increasing renewable energy generation has a negative effect on German day-ahead electricity prices.

To test whether the negative impact of renewable energy sources increases over time, taking into account that renewable energy capacity increases, I test the second hypothesis:

2. H0: The negative impact of renewable energy generation increases over time within the time period 2004-2017.

for the period 2009-2010. If the value of feed-in tariff policies – providing renewable energy producers a guaranteed price for a certain amount of supplied electricity that covers generation costs – is subtracted, savings are still economically significant. O’Mahoney and Denny (2011) also assess the savings resulting from an increasing share of renewables. As Ireland is a closed market that has little cross-border trading, the impact that wind energy has on the Irish electricity market is measured precisely. O’Mahoney and Denny find, using a time-series analysis based on hourly data for 2009, that historic costs would have been 12% higher in the case without any wind energy. Whether the merit order will be the same if feed-in tariffs would be zero, is not researched extensively yet. Hildmann, Ulbig, and Andersson (2015) investigate the assumption that both wind and solar energy have zero marginal costs. They find that conventional power plants still need to exist using true marginal costs of renewable energy sources. These ‘true’ marginal costs consist of wear and tear costs, land-lease costs, a concession tax and a forecasting error where system operators have to cover for. From the perspective of conventional power plants, the day-ahead electricity prices are too low to operate profitably. Hildmann et al. (2015) state that ‘gas-fired plants cannot operate profitably because peak day-ahead prices are too often below their marginal operation costs’. Energy storage facilities face this same problem. A solution may be to invest in flexible gas-powered plants that are able to be shut off when demand can be met by renewable energy sources, as Machiel Mulder proposes in an interview with Het Financieele Dagblad.

3. METHODOLOGY

A side-effect of an increasing share of electricity from renewables is the sensitivity to weather conditions renewables like wind and solar power generally have. When the electricity from renewables has a greater share in the total energy mix, the impact of weather conditions like wind velocity and the daily duration of sunlight are likely to have a more substantial impact on day-ahead electricity prices. In Mulder and Scholtens (2013), these effects are described as the climate factors. The climate factors I use are the wind velocity, sunshine duration and the temperature of river water. Next to the climate factors, Mulder and Scholtens suggest that economic factors influencing the demand and supply of electricity logically also impact electricity prices. To explain movements in the electricity price, I include total electricity generation, total net load, and the marginal cost of production, which is an important factor of the merit-order effect. I elaborate more on these variables in the next section. To adjust for weekly patterns (e.g., low prices in the weekends due to lower demand levels), day of the week dummies are added to the model. Seasonal patterns are provided for by means of the daylight variable. Daily duration of sunshine is higher on summer days than in the winter. In both the summer and winter, electricity prices are generally higher than in the fall and spring due to higher cooling and heating demand in the summer and winter, respectively. These cooling and heating effects result in a nonlinear relation between climate conditions and electricity prices (Bessec and Fouquau, 2008). Therefore, the impact of a growing share of renewables on electricity is likely not linear. To deal with this nonlinearity, I estimate the model in logs, except for the variables sunshine duration, wind velocity and the temperature of river water as they will have values equal to zero within the researched period of time.

Ultimately, the equation I test is

𝐿𝑜𝑔 𝑃 = 𝛽'+ 𝛽)Log 𝑃-./ + 𝛽0Log 𝐿 + 𝛽1Log 𝐺 + 𝛽3(𝑆) + 𝛽7(W) +

𝛽₉ 𝑅𝑊𝑇 + 𝛽₌𝐷𝑢𝑚𝑚𝑦𝑆𝑢𝑛𝑑𝑎𝑦 + ⋯ + 𝛽₎₀𝐷𝑢𝑚𝑚𝑦 𝐹𝑟𝑖𝑑𝑎𝑦 + 𝜀, (1)

where P is the German day-ahead electricity price, Pgas is the German day-ahead

for the period before 2004. The period 2004-2017 shows substantial growth in renewable energy capacity, making this period worthwhile of examination. I isolate five sub-periods within this period to evaluate the change in the impact of the explanatory variables over time. For some of the variables, only data is available for a smaller time-period. In that case, I regress these variables for the available time-period. In the next section, further elaboration on the variables used in equation (1) is provided.

4. DATA

To estimate and explain the movements of the German electricity price, I estimate a model that uses both economic and climate factors as explanatory variables to explain German day-ahead electricity prices. For the explanatory variables, I partly follow Mulder and Scholtens (2013). The economic factors are explained by the total net load, the total electricity generation and German gas prices. The climate factors are explained by the wind velocity, the duration in sunshine and the temperature of river water in Germany. Table 1 displays the descriptive statistics of the variables discussed. Below, I expand on the explanatory variables and the dependent electricity price itself.

Table 1. Descriptive statistics

Mean Std. dev. Min. Max.

Log(Pelectricity) 4.45 0.21 -0.29 6.00 Log(Pgas) 2.97 0.29 1.95 3.68 Log(Net Load) 15.48 0.12 15.12 15.71 Log(Tot. Gen.) 15.57 0.16 12.29 15.91 Sun 4.55 3.63 0.00 15.84 Wind 3.52 1.31 0.00 13.62 RWT 0.05 0.24 0.00 2.45

Pelectricity and Pgas are the daily average German day-ahead electricity price and day-ahead gas price, respectively.

Net Load describes the daily amount in MW that is on the grid. Tot. Gen. represents the daily total electricity generation in Germany.

The variable for the river water temperature (RWT) is measured as the temperature above the threshold of 23 ºC.

Sun and Wind are the daily average sunshine duration in hours and wind velocity in m/s, respectively.

Prices

market for the delivery on the following day. The historical data refers to the day on which the contract was traded. Figure 3 displays the German day-ahead electricity prices for the period 2000-2017. A downward trend is noticeable from 2008 to 2017 which may be explained by the increase in the capacity of renewable energy sources. The graph displays high volatility for the period 2005-2008. Volatility seems to be decreasing after this period which is not in line with the impact a growing share of renewable energy capacity should have on electricity prices. The impact of (volatile) climate conditions should increase volatility due to this increasing share of renewable energy capacity.

Figure 3. The German day-ahead electricity price for the period 2000-2017. Source: Bloomberg Tightness in the market

Total electricity generation

To measure the impact that electricity demand has on the electricity price, I use daily averages of total German electricity generation for the period 2015-2017. Following German policy, electricity from renewable energy sources are generally used first (Klessmann et al., 2008). Therefore, the total generation of electricity will be a reasonable proxy to capture the whole merit-order curve up to the point where demand meets supply. With a high demand, more power plants with relatively high marginal costs are utilized, increasing the equilibrium electricity price. Also, following standard economic theory, an increase in demand automatically increases the electricity price with an upward sloping supply curve. Both result in a positive relationship between demand and electricity prices. The total electricity generation is also derived from the databank of Entsoe.

Marginal cost of production

Since gas-fired-powered capacity has a substantial part in the merit order, day-ahead natural gas prices are used as a proxy for the marginal cost of production. I use daily averages for this variable. At peak-demand, gas-fired power plants are the last plants utilized to meet demand. Therefore, this source is a reliable proxy to cover the marginal costs of production. When gas prices increase, the supply curve shifts upwards resulting in higher electricity prices. Consequently, I expect to find a positive relationship between gas prices and German electricity prices. Gas prices decreased to a minimum of €7 per MWh in 2009 and reached a maximum of nearly €40 in 2012. Nowadays, it decreased back to its mean level of €20.

Wind velocity

five stations spread over Germany (Berlin, Düsseldorf, Hannover, Lübeck, München). Due to limited data availability, I use daily values for this variable. Since I use daily average values for this variable, the observed impact may be lower than actual values. For the period 2000-2017 I find a daily average wind velocity of 3.52 m/s, with an average maximum of 13.62 m/s. As wind turbines are highly impacted by the wind velocity, and wind energy has relatively low marginal costs, I expect to find a negative relationship between wind velocity and electricity prices.

Daylight sunshine duration

Following the merit-order effect, more daylight hours will have a negative impact on the electricity price, since daylight positively impacts the electricity production by solar power. Assuming such an increase will shift electricity supply to the right, prices will decrease. In addition, longer days will lead to less lighting, and thus less electricity usage. This will likely enhance the negative merit-order effect even more as a decrease in demand logically leads to lower prices. I take the daily average of daylight sunshine duration in hours measured at the same five locations as where wind velocity is measured.

River water temperature

The river water temperature may have an impact on the electricity price as well. For environmental reasons, plants are obliged to reduce production when the temperature of the river exceeds a certain threshold. This water is used for cooling of the plant. When the temperature reaches this threshold, supply will decrease from those plants using river water for cooling purposes. When plants produce less, this causes the merit-order curve to shift to the left. Dependent on demand, this logically causes the electricity price to increase. I adopt the threshold used in Mulder and Scholtens (2013). The variable RWT (River Water Temperature) is measured as the temperature of the river above the threshold of 23 ºC. In the period of 2000-2017, the temperature of the river only exceeds the threshold level of 23 ºC with a maximum of 2.45 ºC and with a mean of 0.05 ºC due to a value of zero for this variable in most days. I expect to find a positive relationship between RWT and electricity prices, as higher temperatures result in a leftward shift in the supply curve increasing prices.

stable and secure electricity supply. To assure this, wind and solar energy should be negatively correlated. In other words, if there is a peak in wind energy supply, the supply of solar energy should be low and vice versa. In table 2, we notice a negative correlation of -0.33, which is in line with the research of Solbakken et al. The other variables also show an expected correlation. With respect to electricity prices, the price of gas is positively correlated, which is evidence in favour of the merit order. Both net load and total generation are positively correlated, as standard economic theory of supply and demand suggests. Both the duration of sunshine and wind velocity are negatively correlated to electricity prices. As expected, the duration of sunshine has a lower correlation than wind velocity. This can partly be explained by the relatively lower capacity of solar energy in Germany. The negative correlation between the duration of sunshine and the net load and total generation may be evidence for the impact that sunshine has on electricity usage, as less lighting is needed when days last longer and vice versa. Also, a positive correlation between wind velocity and the net load and total generation confirms the phenomenon that electricity demand is higher when the weather is bad. Lastly, the temperature of the river water has a slightly positive correlation. Due to the variable being zero for most of the observations, the correlation is relatively low.

Table 2. Correlation Matrix Log(Pelectri city)

Log(Pgas) Log(Net

Load) Log(Tot. Gen.) Sun Wind RWT Log(Pelectricity) 1 Log(Pgas) 0.21 1 Log(Net Load) 0.23 0.12 1 Log(Tot. Gen.) 0.16 0.10 0.62 1 Sun -0.07 -0.11 -0.24 -0.23 1 Wind -0.25 0.04 0.12 0.20 -0.33 1 RWT 0.04 0.14 -0.05 -0.07 0.18 -0.03 1 Pelectricity and Pgas are the daily average German day-ahead electricity price and day-ahead gas price, respectively.

Net Load describes the daily amount in MW that is on the grid. Tot. Gen. represents the daily total electricity generation in Germany.

The variable for the river water temperature (RWT) is measured as the temperature above the threshold of 23 ºC.

Statistics

The Dickey-Fuller test is used to test for stationarity. The null-hypothesis of non-stationarity is tested against the alternative of a stationary process. Formally, we write this as H0: 𝜓 = 0, H1: 𝜓 < 1 in the regression: 𝑦_O= 𝜑𝑦_OQ)+ 𝑢_O, where 𝜓 = 𝜑 −

1. After performing the Dickey-Fuller test for all variables, I cannot reject the null-hypothesis and I conclude that all variables are stationary.

To test for heteroscedasticity, I perform a White test on my estimation. The White test tests the null hypothesis of homoscedasticity against the alternative hypothesis of heteroscedasticity. The null hypothesis also states that the errors are independent of the regressors, and that the linear specification of the model is correct. It uses the residuals of the series and regresses its variance against all the explanatory variables and its interaction terms. The null of homoscedasticity is not rejected if all the coefficients of these terms are equal to zero. In contrast, if one or more of these coefficients are different from zero, the null hypothesis can be rejected, and heteroscedasticity is assumed. I apply Heteroscedasticity and Autocorrelation Consistent (HAC) Newey-West Robust standard errors to control for any presence of heteroscedasticity and autocorrelation.

5. RESULTS

Table 3 displays the regression output of the model regressed in several time periods within the period 2004-2017. Five time periods are regressed to enable assessing the change in impact of the explanatory variables over time. Both variables net load and total generation are available for the period 2015-2017 and the gas prices are available for the period 2007-2017. All the other variables are available for the researched time period 2004-2017.

The output shows a significant positive relationship between the log of day-ahead gas prices and the log of electricity prices, suggesting a positive elasticity between those two as expected. This elasticity is the highest in the period 2016-2017 at an elasticity of 0.34. A 1% increase in the gas price results in a 0.34% increase in the electricity price in this period. This is evidence for gas-powered plants to be still the price-setting energy source of the merit order.

with 0.45% in the period 2015-2017. For the period 2016-2017 I do not find a significant result for the net load. For the total electricity generation, I do not find any significant results over the period 2015-2017. However, when leaving the net load out as an explanatory variable, I do find a significant positive elasticity of 0.11 for the total electricity generation in the period 2015-2017. As the net load and total generation are closely correlated – they have a correlation of 0.62 – this explains one of the two becoming insignificant in the regression of table 3.

I find that the duration of sunshine per day has a significant but not an economical impact as its coefficient is equal to -0.01 at its highest impact. I do not find a growth in this value over time, suggesting the impact of solar power on German electricity prices has not grown significantly over the past years. This low impact can be explained by the relatively low capacity of solar energy in Germany. However, when days are longer less lighting is needed, and hence electricity demand is relatively low. This results in a negative impact of sunshine on demand, and therefore on electricity prices. This latter effect seems not to be embedded in the found coefficient. This result however is in line with the paper of Mulder and Scholtens (2013) who do find that the sunshine has no significant impact on electricity prices.

Wind velocity does have a significant negative impact on electricity prices. For the period 2004-2007, I find a coefficient of -0.31, which increases to -0.44 in 2016-2017. The coefficients are significant at the 1% level for all estimated time periods. This result is evidence in favour of the merit-order effect. The results are also in line with the lead paper of Mulder and Scholtens (2013). The increase in the impact that wind velocity has can be explained by the growing capacity of (mostly onshore) wind energy in Germany as shown in figure 1. Onshore wind capacity has grown from 37,757 MW in 2015 to 49,404 MW in 2017. This is equal to a 31% increase over the past three years.

supply decreases. This results in a leftward shift of the merit-order curve which logically increases the electricity price.

Actual renewable generation

To substantiate the results found by regressing equation (1), the actual share of wind and of solar energy generation is used to measure the direct price-effect these renewable energy sources have on German day-ahead electricity prices. Due to limited data availability, I am only able to assess the impact over the period 2015-2017. Ultimately, I regress the following equation similar to (1)

𝐿𝑜𝑔 𝑃 = 𝛽_'+ 𝛽₎Log 𝑃_-./ + 𝛽₀Log 𝐿 + 𝛽₁Log 𝐺 + 𝛽₃(𝑆_STUV) + 𝛽₇(𝑆_/WU) +

𝛽9𝐷𝑢𝑚𝑚𝑦𝑆𝑢𝑛𝑑𝑎𝑦 + ⋯ + 𝛽))𝐷𝑢𝑚𝑚𝑦 𝐹𝑟𝑖𝑑𝑎𝑦 + 𝜀, (2)

where P is again the German day-ahead electricity price, Pgas is the German

day-ahead gas price, L is the total net load on the grid, and G is the total electricity

generation. Swind and Ssun are the share of electricity generation by wind turbines

and solar panels, respectively. In table 4 the regression output of equation (2) is shown for the period 2015-2017. The results of the regression are shown on a yearly basis to assess the change in the coefficients over time. Table 4 also displays the regression for the period 2015-2017 as one time-period. One thing that catches the eye is that the coefficient for the log of gas prices is insignificant for both years 2015 and 2017. This implies that a change in gas prices in these years had no significant impact on the electricity price. It may be explained by coal-powered plants being the price-setting plants for most of the observations in these periods instead of gas-powered plants. Sensfuß et al. (2008) suggest that both gas- and coal-powered plants have a price-setting role in the electricity market. Overall, a significant elasticity of 0.18 is found for gas prices over the period 2015-2017 which is almost the same as the 0.19 found for the regression of equation (1).

Similar to the regression of equation (1), I find a positive elasticity for the net load. For the periods 2015 and 2015-2017 I find an elasticity of 0.49 and 0.47, respectively. Both are significant at the 1% significance level. Where the elasticity for net load is low, the elasticity of total electricity generation is relatively high. This is likely due to the correlation of the net load and the total electricity generation of 0.62, as discussed in the results of equation (1). Overall, I do not find any substantial differences for the values for the net load and total electricity generation between the regression of equation (1) and (2).

2016 and -1.05 in 2017. The coefficient for the period 2015-2017 is -0.71. An increase in the share of wind energy of one percentage point results in an average price decrease of 0.71%. To illustrate, for an average increase of 9350 MW, which is 1% of the daily average wind energy generation over the period 2015-2017, the electricity price decreases with 0.71%. For solar power, I also find an overall significant negative impact for the period 2015-2017. Only in 2017 the coefficient is not significant at the 10% level. Therefore, I can only conclude that the impact is growing between 2015 and 2016 with a growth of 0.19. When the share of solar power generation increases with one percentage point, daily average electricity prices decrease 0.19% more in 2016 than in the previous year.

Table 4. Regression output of the share of renewables on the log of the daily average day-ahead electricity price 2015 2016 2017 2015-2017 Log(Pgas) -0.05 (0.05) 0.36 (0.06)*** 0.35 (0.26) 0.18 (0.03)*** Log(Net Load) 0.49 (0.08)*** 0.20 (0.12) 0.04 (0.23) 0.47 (0.10)*** Log(Tot. Gen.) -0.05 (0.04) 0.04 (0.02)** 0.40 (0.19)** 0.00 (0.03) Share wind -0.63 (0.05)*** -0.66 (0.11)*** -1.05 (0.31)*** -0.71 (0.14)*** Share sun -0.35 (0.14)** -0.54 (0.18)*** -0.63 (0.66) -0.60 (0.14)** Dummy Monday 0.11 (0.02)*** 0.14 (0.03)*** 0.22 (0.10)** 0.13 (0.04)*** Dummy Tuesday 0.10 (0.02)*** 0.10 (0.03)*** 0.18 (0.11)* 0.11 (0.04)*** Dummy Wednesday 0.09 (0.02)*** 0.12 (0.03)*** 0.18 (0.12) 0.10 (0.04)** Dummy Thursday 0.08 (0.02)*** 0.10 (0.03)*** 0.19 (0.11) 0.10 (0.04)** Dummy Friday -0.01 (0.02) 0.04 (0.03) 0.09 (0.10) 0.02 (0.03) Dummy Sunday 0.23 (0.02)*** 0.18 (0.02)*** 0.28 (0.09)*** 0.24 (0.03)*** Variance equation Constant -2.12 (1.13)* -0.31 (1.79) -3.33 (2.32) -3.43 (1.48)** Adjusted R2 0.61 0.54 0.23 0.32 No. observations 365 366 335 1066

Standard errors in parentheses. *,**,*** refer to 10%, 5% and 1% significance levels, respectively.

Pgas is the daily average German day-ahead gas price.

Net Load describes the daily amount in MW that is on the grid. Tot. Gen. represents the daily total electricity generation in Germany.

Share wind and Share wind are the % of total generation for wind energy generation and solar energy generation, respectively.

6. CONCLUSION

consequence of the increasing capacity of wind energy in Germany as shown in figure 1 and 2. Moreover, I provide evidence in favour of the merit-order effect by showing that there exists a negative relationship between electricity prices and both solar and wind energy generation and by showing a positive elasticity between gas and electricity prices. I can conclude that the merit-order effect is stronger than the positive impact an increasing share of renewables may have on electricity prices due to de-investments, balancing costs and additional transmission costs. To give an answer to my research questions and to conclude on my hypotheses, I present the hypotheses below.

1. H0: An increasing renewable energy generation has a negative effect on German day-ahead electricity prices.

A negative relationship is found between the wind velocity and German electricity prices. Moreover, I find that actual wind energy generation has a significant negative impact on day-ahead electricity prices. However, the impact of sunshine duration is not economically significant. An explanation could be that the growth in German solar capacity is relatively low compared to the growth of wind capacity. When I regress actual solar energy generation on electricity prices, I find a slightly lower than wind energy, but significant negative impact in the period 2015-2017. 2. H0: The negative impact of renewable energy generation increases over time

within the time period 2004-2017.

daily data. Additionally, I do not take cross-border trade into account in this paper. As the electricity market in Germany is connected with several countries in Europe, the ability to include import and export structure would have given more realistic and precise outcomes. The results have substantial implications for the electricity market as a whole, and for conventional power plant operators who are dealing with decreasing and fluctuating electricity prices. Firms that cannot operate profitably because peak prices do not cover their marginal operation costs anymore do need to find solutions to stay operative. On the other hand, renewable energy sources do not offer a stable electricity price yet, which makes the conventional power plant indispensable for the time being. This leaves an opportunity for conventional power plants to invest in flexible power plants that have to be able to be turned off when renewable energy sources are capable to meet demand (Grol, 2018). From an investor-perspective, this paper draws a substantial implication as well. With an increasing impact of weather conditions on the electricity market, hedging using financial instruments like weather derivatives is likely to become more effective in the future. This is an interesting topic for future research. Other suggestions are to include the effect that an increase in renewable energy production has on the import and export structure of electricity. Furthermore, not all costs of various power plants are considered. Although this would be challenging to test using daily data, it will provide a more appropriate valuation of the relationships tested in this paper. Fourth, high investment costs of renewable energy production are not accounted for in this paper as I assume that renewable energy sources are at the left of the merit order. If these costs are taken into account, the merit order may even change its order (Hildmann et al., 2015). As the costs of renewables are subsidized by the government for a substantial part, the impact of policy on the electricity market is an interesting field to explore as well.

7. REFERENCES

Axthelm, W. (2018, Jan 25) Ausbauzahlen für das gesamtjahr 2017 in Deutschland windenergie an land: Starker zubaupfad im übergangsjahr, EEG

reparieren und klimaschutz stärken. Retrieved from

https://www.wind-

Bessec, M., & Fouquau, J. (2008). The non-linear link between electricity consumption and temperature in Europe: a threshold panel approach. Energy Economics 30(5), 2705-2721.

Bley, S. J., Hametner, M., Dimitrova, A., Ruech, R., De Rocchi, A., Gschwend, E., & Umpfenbach, K. (2017). Smarter, greener, more inclusive? Indicators to support the Europe 2020 strategy. Eurostat statistical books 2017 edition, 81-106.

Cludius, J., Hermann, H., Matthes, F. C., & Graichen, V., 2014. The merit order effect of wind and photovoltaic electricity generation in Germany 2008–2016: Estimation and distributional implications. Energy Economics 44, 302-313.

Grol, C. (2018, May 24). Er moeten meer gascentrales komen en ze zullen

rendabel zijn. Retrieved from

https://fd.nl/economie-politiek/1255208/er-moeten-meer-gascentrales-komen-en-ze-zullen-rendabel-zijn

He, Y., Hildmann, M., Herzog, F., & Andersson, G., 2013. Modeling the merit order curve of the European energy exchange power market in Germany. IEEE Transactions on Power Systems 28(3), 3155-3164.

Hildmann, M., Ulbig, A., & Andersson, G., 2015. Revisiting the merit-order effect of renewable energy sources. IEEE Power & Energy Society General Meeting.

Klessmann, C., Nabe, C., & Burges, K., 2008. Pros and cons of exposing renewables to electricity market risks—A comparison of the market integration approaches in Germany, Spain, and the UK. Energy Policy 36(10), 3646-3661.

McConnell, D., Hearps, P., Eales, D., Sandiford, M., Dunn, R., Wright, M., & Bateman, L., 2013. Retrospective modeling of the merit-order effect on wholesale electricity prices from distributed photovoltaic generation in the Australian National Electricity Market. Energy Policy 58, 17-27.

Mulder, M., Scholtens, B., 2013. The impact of renewable energy on electricity prices in the Netherlands. Renewable energy 57, 94-100.

OECD/IEA (2014, Oct 8) Renewable energy act, IEA Publishing. Retrieved from

https://www.iea.org/policiesandmeasures/pams/germany/name-21295-en.php

O'Mahoney, A., Denny, E., 2011. The merit order effect of wind generation on the Irish electricity market. Changing Roles of Industry, Government and

Research, 30th USAEE/IAEE North American Conference. International Association for Energy Economics.

Ozdemir, O., Koutstaal, P.R., & Hout, M.V., 2017. Integration costs and market value of variable renewables: A study for the Dutch power market. Policy Studies, Energy Research Centre of the Netherlands.

Sensfuß, F., 2007. Assessment of the impact of renewable electricity generation on the German electricity sector—an agent-based simulation approach. Dissertation, Universität Karlsruhe (TH), Karlsruhe. VDI, Düsseldorf.

Sensfuß, F., Ragwitz, M., & Genoese, M., 2008. The merit-order effect: A detailed analysis of the price effect of renewable electricity generation on spot market prices in Germany. Energy policy 36(8), 3086-3094.

Solbakken, K., Babar, B., & Boström, T., 2016. Correlation of wind and solar power in high-latitude arctic areas in Northern Norway and Svalbard. Renewable Energy and Environmental Sustainability 42, 1-4.

Von Roon, S., Huck, M., 2010. Merit order des kraftwerksparks. Forschungsstelle für Energiewirtschaft e.V.,1-5.

Data Sources:

Actual net load. Retrieved Apr. 30, 2018 from

https://transparency.entsoe.eu/load-domain/r2/totalLoadR2/show

Bloomberg L.P. (2018) Day-ahead electricity price graph for Germany EEX 14/6/2000 to 31/12/17. Retrieved Apr. 30, 2018 from Bloomberg database.

Bloomberg L.P. (2018) Germany Gaspool (GPL) Natural Gas Forward 03/9/2007 to 31/12/17. Retrieved Apr. 30, 2018 from Bloomberg database.

River water temperature. Retrieved Apr. 30, 2018 from https://www.gkd. bayern.de/fluesse/wassertemperatur/download/index.php?thema=gkd&rubrik=flu

esse&produkt=wassertemperatur&msnr=783500113&gknr=0&beginn=&ende

Total electricity generation. Retrieved Apr. 30, 2018 from